Professional Documents
Culture Documents
Application of Reference
Architectures for
Enterprise Integration
7.1 Introduction
7.2 Enterprise Reference Architectures
7.3 CIMOSA
Introduction • CIMOSA Approach • CIMOSA Modeling
Framework • CIMOSA Integrating Infrastructure
7.4 Application of CIMOSA at Traub
Introduction • Reorganization Objective • Requirements
Definition • Architectural Design • Detailed
Design • Implementation • The Role of CIMOSA in Traub’s
Reorganization Process
7.5 The Practical Value of CIMOSA
Arian Zwegers
7.6 Two Other Enterprise Reference Architectures
Baan Development
GRAI/GIM • PERA
Henk Jan Pels 7.7 Baan’s Enterprise Modeler
Eindhoven University of Technology 7.8 Summary
7.1 Introduction
Enterprises face integration problems due to necessary internal adaptations to cope with severe competition
in the global marketplace. To survive at the global market, efficient operation and innovative management
of change are essential. Enterprises need to evolve and be reactive, so that change and adaptation should
be a natural dynamic state rather than something occasionally forced onto the enterprise. Enterprise
engineering is the discipline that identifies the need for change in enterprises and carries out that change
expediently and professionally. Enterprise engineering intends to cater for continuous evolution and to
achieve a certain level of enterprise integration. Enterprise integration propagates the integration of all
enterprise resources to optimize business operations. That is, enterprise integration involves the collec-
tion, reduction, storage, and use of information, the coordination of product flows, the organization of
machines, the integration of human actions, etc.
Recently, some reference architectures have been defined that aim to provide the necessary frameworks
for enterprise integration. They are based on the idea that large parts of enterprise integration projects
are similar in every type of enterprise. These parts could be standardized and supported by methodologies,
tools, and other products that facilitate integration projects (Bernus et al., 1996).
The objective of this chapter is to investigate the practical value of enterprise reference architectures
concerning the integration of production systems. For this, three enterprise reference architectures have
Wyns et al. (1996) state that the benefits of reference architectures lie, among others, in unified and
unambiguous terminology, in envisaged simplicity of designing specific systems, and in high quality by
relying on proven concepts. In addition, a reference architecture models the whole life history of an
enterprise integration project. It indicates and justifies how and at what stage in the development process
external constraints and engineering design decisions are introduced, thereby providing traceability
between solution-independent requirements and final realizations. It provides the framework in which
enterprise integration methodologies work.
The IFAC/IFIP Task Force on Architectures for Enterprise Integration has developed an overall frame
work to organize existing enterprise integration knowledge. The proposed framework was entitled
GERAM—Generalized (or Generic) Enterprise Reference Architecture and Methodology. GERAM is
about those methods, models, and tools that are needed to build and maintain the integrated enterprise.
GERAM defines a toolkit of concepts for designing and maintaining enterprises for their entire life cycle.
The scope of GERAM encompasses all the knowledge needed for enterprise engineering. Thus, GERAM
provides a generalized framework for describing the components needed in all types of enterprise
engineering processes. GERAM provides a description of all elements recommended in enterprise
engineering and integration (IFAC/IFIP, 1997).
Fig. 7.1 shows the components of the GERAM framework. The most important component is the
Enterprise Reference Architecture, which defines enterprise-related concepts recommended for use
in enterprise engineering projects. The reference architecture should point toward purposeful organization
of enterprise concepts (Nell, 1996); it should identify and structure concepts for enterprise integration,
most notably life-cycle activities and architectural concepts such as abstraction, hierarchy, and views.
GERAM distinguishes between the methodologies for enterprise engineering and the modeling languages
that are used by the methodologies to describe and specify the structure, content, and behavior of the
enterprise. These languages will enable the modeling of the human part in the enterprise operation as
well as the part of business processes and supporting technologies. The modeling process is supported
by guidelines, and the results of the process are enterprise models that represent all or part of the enterprise
operations, including its manufacturing or service tasks, its organization and management, and its control
and information systems. These models should be used to improve the ability of the enterprise to evaluate
operational or organizational alternatives (e.g., by simulation), and thereby improve its current and future
performance.
The methodologies and the languages used for enterprise modeling are supported by enterprise
modeling tools. The modeling process is enhanced by using reference models that provide reusable models
of human roles, processes, and technologies. The operational use of enterprise models is supported by
specific modules that provide prefabricated products such as human skill profiles for specific professions,
common business procedures (e.g., banking and tax rules), or IT infrastructure services.
7.3 CIMOSA
Introduction
The goal of the ESPRIT project AMICE (reverse acronym of European Computer Integrated Manufacturing
Architecture) was to develop an Open System Architecture for CIM (CIMOSA). The project started in
the mid-1980s and finished after some extensions in the mid-1990s. CIMOSA should facilitate continuous
enterprise evolution and make necessary operational improvements manageable. At the same time,
CIMOSA should provide a strategy to deal with legacy systems to protect current and planned investment
in an enterprise’s production process. CIMOSA aims to offer support for enterprise integration, more
precisely for “business integration” and “application integration” (AMICE, 1993a).
Business integration is concerned with the coordination of operational and control processes to satisfy
business objectives. In every enterprise, processes are present that provide supervisory control of the
operational processes and coordinate the every-day execution of activities. CIMOSA aims to integrate these
processes by process-oriented modeling of enterprises. Modeling these processes and their interrelations
can be used in decisions regarding the requested level of business integration.
Application integration, which affects the control of applications, is concerned with the usage of
information technology to provide interoperation between manufacturing resources. Cooperation
between humans, machines, and software programs must be established by the supply of information
through inter- and intra-system communication. CIMOSA tries to support integration at this level by
defining a sufficient infrastructure to permit system-wide access to all relevant information regardless
of where the data reside.
In addition to business and application integration, a third level of integration is discerned by the
AMICE project, namely physical system integration. This level is concerned with the interconnection of
physical systems and has led to a number of standards, such as OSI and MAP. CIMOSA supports this
type of integration by adherence to the standards.
CIMOSA Approach
CIMOSA provides a framework that guides designers in the design and implementation of CIM systems.
In addition, it aims to guide vendors in the development of CIM system components, so that these
components can be added and removed at will. It does not provide a standard architecture to be used
by the entire manufacturing industry, but rather a reference architecture from which a particular
enterprise can derive particular architectures that fulfill its needs. As such, CIMOSA adheres to a
descriptive, rather than a prescriptive methodology (AMICE, 1993a). Fig. 7.2 gives an overview of the
CIMOSA concepts.
The CIMOSA modeling framework enables one to describe a particular enterprise. Accompanying
methods, called “model creation processes,” guide engineers in the creation and maintenance process to
obtain and maintain a consistent system description. CIMOSA supports the explicit description of
enterprise processes at different levels of abstraction for strategic, tactical, and operational decision-
making. Simulation of alternatives and evaluation of design decisions enhances the decision support.
CIMOSA supports incremental modeling of the enterprise rather than following an overall top-down
approach. In short, CIMOSA allows the end user to define, prototype, design, implement, and execute
its business processes according to his/her needs.
CIMOSA facilitates a system life cycle that guides the user through model engineering and model
execution. The life cycle starts with the collection of business requirements in a requirements definition
model, and goes, through the translation of the requirements into a system design model, to the
description of the implemented system. These phases are followed by a model release for operation and
model execution for operational control and monitoring. However, various methodologies consisting
of various system life-cycle phases are possible to instantiate particular models from the reference
architecture. These methodologies are supported by tool sets, which are defined by enterprise engineering
implementation models.
Model engineering and model execution are supported by the CIMOSA integrating infrastructure.
This infrastructure provides a set of generic services that process the released implementation model,
provide access to information, and connect to resources. In addition, the integrating infrastructure hides
the heterogeneity of the underlying manufacturing and information technology.
In the next two subsections, the two most important parts of CIMOSA—namely, the CIMOSA
modeling framework and the CIMOSA integrating infrastructure—are explained in more detail.
Instantiation
Requirements
Definition
Derivation
Design
Specification
Introduction
The ESPRIT project VOICE (Validating OSA in Industrial CIM Environments) validated the CIMOSA
results by the development of manufacturing control and monitoring solutions according to the CIMOSA
concepts in three types of industries. The VOICE project consisted of vendors, system integrators, research
institutes, and industrial partners. In each of the three industrial pilots, a different aspect of the manu-
facturing process was addressed. This section focuses on one of these pilots, namely the Traub pilot.
Traub’s motivation regarding the validation of CIMOSA was to examine whether CIMOSA is a useful
tool to support the restructuring process of the existing manufacturing organization and to support the
new organization with implemented information technology (Gransier and Schönewolf, 1995). Besides
some mentioned references, the main sources of this section are VOICE meetings and reports.
Traub AG is a German manufacturer of machine tools. The products cover conventional machines as
well as NC machines for turning and milling. The economic downturn in the market and the world-wide
recession have forced all machine tool manufacturers to improve their cost situation and shorten the
terms of delivery. Customer demands are characterized by an increased diversification of products,
resulting in decreasing numbers of series production. Enormous sales efforts and the introduction of
new products are not the only guarantee for survival. Enhanced demands to become more flexible and
efficient forced Traub to reorganize its production department (Schlotz and Röck, 1995).
Figure 7.5 shows Traub’s production control system before the reorganization. At that time, it consisted
of a mainframe with the Production Planning System, an IBM RS/6000 for area control functionalities,
and several cell controllers. Traub undertook global production planning for a year and a half in advance.
In this planning, machine types, options, and number of products were incorporated. Every 10 days, a
partial planning was made, which consisted of timed orders for design, purchase, and production control.
It was possible to control the order flow from the area controller by releasing orders, standing in a “10 days
order pool.” Dedicated applications established the transmission of NC-programs between the area
controller and the cell controllers. The worker at the NC-machine obtained a list of order data, and could
transfer and edit NC-programs from and to the machines, and transmit the optimized programs to the
area controller. Order progress and other monitoring data could be automatically sent from the machines
to the area controller or by entering a message on a terminal. Monitoring information was sent to the
mainframe at the production planning level from terminals on the shop floor located at several places
near the machines.
One of the problems associated with the old situation at Traub was its flexibility to respond to changes
in customer needs. Shorter delivery times with simultaneous reduction of stocks were already forcing
Traub to shorten machining times and to link all processes in design, planning, and shop floor more
closely. Even more, it frequently became necessary to rearrange plans at short notice because of changes
in demand, machine downtimes, missing production facilities, urgent contracts with large customers,
absence of staff, or rejects. Rescheduling of manufacturing jobs became difficult due to the limited
feedback from the shop floor, and expenditures for short-notice re-work and machine modifications
increased sharply.
In particular, the preparatory sectors were the areas where increased deadline pressure and the tendency
toward smaller and smaller batch sizes added to planning complexity and expenditure. Most notably, tool
management became problematic because the lack of up-to-date information concerning the availability
of machine tools and their components also reduced the production capacity while operators waited for
resources. Furthermore, production management had to cope with a growing amount of specialized tools
necessary to produce individually designed machine parts. There were nine machine centers, that each
Other applications:
- CAD
- Purchase NC Programming
Token Ring
IBM RS6000
Ethernet
Terminal
Transport Transport
NC-machine NC-machine
having a stock of approximately 120 tools. Each tool consists of eight to ten components. At that time,
there was no registration of which tool or component was in which machine (Schönewolf et al., 1992).
Reorganization Objective
The most important target of the reorganization was to optimize the order throughput time. To achieve
this, a number of changes had to be implemented related to the cooperation on optimized order scheduling
between the area controller and the cell controllers. Other changes concern tool management, monitoring
data acquisition, etc. In this sub-section, only a few major changes are discussed.
The old, process-oriented manufacturing departments such as turning, milling, and grinding were
changed for a more flexible organization. The new organization is based on the principles of cellular
manufacturing, which involves the creation of clusters of machines that are designed and arranged to
produce a specific group of component parts (Black, 1983). In particular, Traub expected the cellular
manufacturing principles to result in a decrease in setup times.
An area control system was required that had to perform the fine planning of manufacturing device
utilization under rescheduling conditions with the help of a production scheduling simulation tool. This
area control system had to be linked enterprisewide with existing production planning, CAD, and CAM
applications. It had to control not only the manufacturing processes, but also the delivery of material,
tools, and equipment. Furthermore, it had to take into account the actual status of each NC-machine.
This information had to be transferred online—without any influence of the working people—directly
from the machine controller to the planning application of the area controller.
Fine-planning had to be supported by efficient tool management to get a tool-optimized order sequence
to decrease the setup time for each machine. This system would provide information on the location
Requirements Definition
Traub defined requirements in a functional sense for the application to be developed, and in a technological/
physical sense for the application’s underlying infrastructure. In other words, requirements were defined
for both the functional and technology aspects of the various components in the manufacturing system.
In addition, requirements were imposed on the reengineering process, influenced by financial and time
aspects.
Traub described its requirements from a functional point of view in terms of a scenario. In this scenario,
the needed functions and information flows, and their places in the total manufacturing system, were
outlined. For example, part of the scenario for the fine-planning activity (precision planning) was
described as follows.
“The work sequences released are assigned to the individual machines during the automatic precision
planning. During a planning sequence, only the machine capacity is taken into account. Starting from the
terminating time given by the order pool system, timing is done backwards. If the calculated starting time
lies in the past, the timing is done forwards while considering the transition time. If the given terminating
time is not observed during this forward scheduling, this is explicitly displayed in the system. No alternative
machine groups are considered in the automatic precision planning” (Schönewolf et al., 1992).
Requirements for the technology to be used were given as well. Besides Traub’s current needs, requirements
took into consideration the existing manufacturing system, strategic aspects such as standards, the factory
environment, and production process constraints. Traub mainly defined its requirements of the infra-
structure in terms of the CIMOSA integrating infrastructure; for example:
• “Multiple machines must be connected to a homogeneous system (presentation services).
• Connectivity to multiple databases on heterogeneous networks must be achieved from different
kinds of computer systems (information services).
• The network must be transparent (common services)” (Schönewolf et al., 1992).
Architectural Design
After requirements were defined, the design of the CIM system commenced. The design activities of
Traub’s reorganization project could be distributed over two phases: architectural design and detailed
design. In the architectural design phase, the system’s functional and technology architectures were
defined, supported by the CIMOSA framework. In the detailed design phase, the system was worked out
in more detail, based on the defined architectures.
Area control was positioned between the production planning level (not in the scope of the reorganization
project) and the shop floor. As an autonomous decision center in a distributed order planning structure,
it processes the order pool coming from the production planning system, and performs scheduling and
management tasks for machine-level planning. The incoming orders are scheduled for a 10-day period
on the various machine groups. For each machine group, the area controller performs a daily planning
to achieve an optimized order schedule with time slices of 2 days. The new order sequence on a machine
is calculated on the basis of the available tools and the list of needed tools, which is extracted from the
NC programs for the new orders. Tools available at NC machines or in the tool store can be requested.
Tool handling is supported by the tool management process.
Being intermediate between production planning and shop floor, area control not only has to provide
the shop floor with orders, but also allows feedback of information from the shop floor to the production
planning system. Then, this system can use data that reflects the actual situation in the shop floor to
optimize the planning process. Based on online messages from the cell control level, the area controller
also supports processing and visualization of data such as the actual status of orders.
Area
Purchase DNC-Node
Control
Non-CIMOSA
Domain
NC-Machine Exception
Handling Handling Domain
Relationship
FIGURE 7.6 Overview of Traub’s production control functions. (Reprinted from Zwegers et al., Evaluation of
Architecture Design with CIMOSA, Figure 2, Elsevier Science, 1997. With permission.)
rder
Pr
Tool
ep
ew O
tus
ar
s Co
r Sta
e
Ne
ess N
w
nsig
Orde
Or
de
Proc
ned
r
CIMOSA CIMOSA
DOMAIN Process Order Tool Data Prepare DOMAIN
Cell Monitoring Monitoring Processing Processing Tools Tool
Control Data Management
Start Man
ls
oo
ols
M
lT
ac
To
hi
tal
ufac
ne
t
Ins
tpu
Da
turing
Ou
t
a
CIMOSA
Manu- Set Up DOMAIN
= Domain Process
facturing NC-Machine NC-Machine
Handling
FIGURE 7.7 Partial refinement of Traub’s domains. (Source: From Zwegers et al., Evaluation of Architecture Design
with CIMOSA, Figure 3, Elsevier Science, 1997. With permission.)
processing of tool data and the preparation of tools. These processes are represented in domain “Tool
Management” by two independent domain processes, namely “Tool Data Processing” and “Prepare Tools.”
Note that the domain processes in Fig. 7.7 influence each other’s behavior.
Domain processes are the root of the decomposition tree. They employ business processes that are
in the middle of the tree. The leaves are called enterprise activities and are employed by business
processes or, if there are no business processes, by domain processes. The behavior of a certain business or
domain process is defined by rules, according to which enterprise activities belonging to this process
are carried out. Enterprise activities represent the enterprise functionality as elementary tasks, and
they are defined by their inputs, outputs, function, and required capabilities. Fig. 7.8 presents a
part of the behavior of domain process “Order Processing.” For clarity, some relationships have been
deleted. Events, which are either received from or sent to other domain processes, are represented by
a Z-shape; enterprise activities are shown as boxes labeled “EA.” Note that there are no business
processes specified for this domain process. The large triangles indicate behavioral rules according
to which enterprise activities or business processes are carried out. Fig. 7.8 was constructed using the
GtVOICE tool.
Note that the requirements definition level of the CIMOSA modeling framework is used at architectural
design activities and not during the requirements definition process. The reason is that CIMOSA does
not support a “true” definition of requirements in the sense as described in the previous subsection.
Instead, the CIMOSA requirements definition level offers support in structuring a manufacturing system,
that is, in defining a functional architecture.
Along with defining a functional architecture, Traub also outlined a technology architecture that
was influenced by the existing infrastructure. When defining functional components, one immediately
maps these components on technological ones; the functional architecture is depicted on the technology
architecture. For example, when a designer defines a control function, he/she decides to execute this
function by a certain type of software module. In addition, the designer determines the interaction of
this component with other technology modules.
Together with its partner, FhG-IPK (Fraunhofer-Institut für Produktionsanlagen und Konstruktion-
stechnik) in Berlin, Traub built a testbed for demonstrating the viability of the new production concepts.
Traub initially defined three production management domains that it considered in its manufacturing
system (or rather in its testbed), namely tool management, order planning (“cell control” in Fig. 7.6),
and machine handling. Subsequently, Traub distributed these three functions over an area controller, a
cell controller, and attached machines.
The technology architecture of the testbed is shown in Fig. 7.9. The area control level comprised
a DecStation 5100 for long-term order planning and tool management. The area controller’s user
interface was implemented via the X.11 protocol on a terminal that was connected to the area
controller via TCP/IP on Ethernet. Communication with the cell controller was established via MMS.
The cell controller was a 486 PC running OS/2, enabling cell controller applications to run in a
multi-tasking environment. A MAP network connected the cell controller with shop-floor devices.
The cell controller received orders from the area controller, after which it processed the orders,
controlling the shop-floor devices via MMS. These devices consisted of a robot controller connected
to an industrial robot, a Traub NC-machine, and a PLC that controlled a conveyor and a round-
table. Whereas the PLC had a MAP interface and connected directly to the MAP network, the NC-
machine and the robot controller communicated with the network via a “protocol converter.” The
converter presented functionalities of both the NC-machine and the robot controller as MMS virtual
machine devices to the MAP network. The machine and the robot controller connected to the
protocol converter via V.24/LSV2.
MMS X.11
Ethernet MMS
Cell Controller
MMS MAP
MMS
Protocol Converter
PLC
V24/LSV2
Drilling Unit
Round Table
Robot Controller
NC-Machine Robot
Detailed Design
In the second design phase, Traub specified its production control system in more detail, elaborating on
the architectures that were defined in the architectural design phase. Both architecture definitions were
further decomposed and worked out into detailed system specifications.
By means of the CIMOSA design specification level, a designer is able to detail a manufacturing system’s
functionality, taking the model at requirements level as the starting point. The requirements definition
modeling level supports a user in the definition of his/her system’s functional architecture, whereas the
role of the design level is to restructure, detail, and optimize the required functionality in a consistent
model. System optimization can be supported by simulation, taking all business and technical constraints
into account. By specifying a model at the design level, a designer describes the full functionality of a
CIM system, while staying within the definition of the functional architecture. If the model at the design
specification level reveals inconsistencies, or the model lacks optimization, it might be necessary to adjust
the model at the requirements definition level.
In addition to detailing the functionality of the system, the designer specifies the technology to be
employed to achieve the required system functionality. Simply stated, CIMOSA prescribes that for each
of the most detailed specified functions, called functional operations, a specified resource is assigned that
provides the required capabilities. The prime task of the design specification modeling level is to establish
a set of (logical) resources that together provide the total set of required capabilities. Some of these
required capabilities are offered by the generic services of the integrating infrastructure.
For example, Traub made a model at the design specification level, elaborating on the previously made
model at the requirements definition level. Traub specified the required information structure, it defined its
most elementary functions, and it assigned resources to these functions. During the creation of the models, the
function and information views were used extensively, whereas the resource and organization views
assigned to
(1,n)
(1,1) described (1,n)
Material Workpiece Work Schedule
by
(1,n)
contains
Entity
(0,n)
Relation
(1,n) set into (1,n)
Program
Subset Relation operation
FIGURE 7.10 Model at particular design specification level, information view (partial). (Source: From Zwegers
et al., Evaluation of Architecture Design with CIMOSA, Figure 5, Elsevier Science, 1997. With Permission.)
were barely addressed. It was not sensible to consider other factors such as responsibility (specified
by constructs of the organization view) before Traub was satisfied with the new control structure
regarding functionality and information flow. Part of the specified information structure is depicted
in Fig. 7.10.
A refinement of specified functions is accompanied by a refinement of the technology that provides the
desired functions. An enterprise looks for adequate products, either commercial ones or user developed.
The technology components, which are defined in the architectural design phase, are specified completely.
The definitions of the CIMOSA integrating infrastructure support this specification. Then, the products that
fulfill the design specifications are selected, considering enterprise policies and constraints. CIMOSA
states that the final build/buy decisions should result in a model that describes the implemented system.
Appropriately, CIMOSA calls this model the implementation description model. However, because the
AMICE consortium defined this part of the modeling framework after Traub reorganized its production
department, Traub does not have any experience with it.
To implement a prototype, Traub identified candidates that might help to implement software modules
or that might be used as complete software modules within the testbed. From the system specification
and the analysis of possible candidates, Traub defined products, toolkits, and tools to implement the
functionalities of the area controller and the cell controller. Furthermore, network interfaces were adopted
and products fulfilling integrating infrastructure services were chosen. For example, Oracle 6 was selected
as the product that had to provide the information services. It was connected by an SQL gateway to
EasyMAP, which was used to provide the testbed’s communication services. The communication services
are part of the common services. In a later stage, FhG-IPK’s communication platform was chosen in
order to offer the desired communication services to the operational system.
Implementation
The final phase contains the implementation and release for operation of the specified CIM system.
Implementation is based on the results and decisions of the previous phases.
Implementation activities comprise those tasks needed to bring the system into operation. During the
implementation phase, Traub procured and built the necessary new physical components. Note that the
word “physical” should not be taken too literally; in software engineering, for example, there are no such
things as “physical components.” The components were tested on the testbed and the correctness of the
underlying logical model was verified. Traub decided to implement its physical system in an evolutionary
Ethernet
Cell Controller 1 Cell Controller 2 Cell Controller 3 Cell Controller 4 Cell Controller 5
Machine Group Machine Group Machine Group Machine Group Machine Group
Scharmann Scharmann 2 Heller Heller 2 Bˆ hringer
TND 30 /42 TND 30 /42 TND 30 /42 TND 30 /42 TND 30 /42
FIGURE 7.11 Technology architecture of Traub’s implemented production control system. (Source: From
Schlotz, C., and Röck, M., Reorganization of a Production Department According to the CIMOSA Concepts, Figure 7,
Elsevier Science, 1995. With permission.)
way, with stable intermediate forms. In this way, Traub hoped to achieve a stable implementation of all
products and their interactions. When the system passed the tests with satisfactory results, the system
was prepared for transfer to production.
Traub felt that user acceptance was a major concern. It realized that the complex changes of the
manufacturing system had to be put carefully to the operators in the production environment. Therefore,
after testing single components as prototypes in the test environment, training sessions for the users of
the area controller and the cell controller were held. To make the workers familiar with the new system,
parts were transferred to the production without being integrated in the existing system. In addition,
the graphical user interface was adapted according to users’ needs, which was done to get a greater
acceptance. When Traub believed that the operators were ready to use the new technologies, the new
components were integrated in the existing system and released for operation.
Finally, the accepted production control system was released for operation. Figure 7.11 shows the
technology architecture of Traub’s production control system as implemented during the reorganization
project.
GRAI/GIM
The GRAI laboratory of the University of Bordeaux, France, has been rather active in the field of enterprise
reference architectures. Along with developing its own ideas, the GRAI laboratory has contributed to the
ESPRIT projects IMPACS and AMICE. Here, the main elements of what has become known as the GRAI
Integrated Methodology (GIM) are described: namely, a “global model,” a modeling framework, and a
structured approach to guide the application of the methodology.
The global model describes the invariant parts of a CIM system: the subsystems, their relationships,
and their behavior. The global model (sometimes called the “macro reference model” or “reference
model”) is based on the concepts of three activity types and their corresponding executional subsystems.
These three subsystems are:
1. Physical subsystem, which performs the activities of product transformation using human and
technical resources
2. Decisional subsystem, which guides production toward its goals
3. Informational subsystem, which feeds the other subsystems with information
Sometimes, a fourth subsystem is distinguished, namely the functional subsystem (Doumeingts et al.,
1987; 1992).
The modeling framework uses IDEF0 formalisms to model the physical and functional subsystems,
GRAI formalisms for the decisional subsystem, and MERISE formalisms for the informational subsystem.
The GRAI formalisms are described below; for the other formalisms, the reader is referred to (Doumeingts
et al., 1995). The GRAI formalisms consist of the GRAI grid and the GRAI nets. The GRAI grid allows
one to model a decision system. It is displayed as a table-like frame, and it uses a functional criterion to
identify production management functions and a decision-cycle criterion to identify decisional levels.
Existing User
System Requirements
Initialization
Detection of
Analysis
inconsistencies
Phase
Consolidated user
requirements
User-oriented
User-oriented
specifications
design
Technical-oriented
Technical-oriented
specifications
design
Implementation
New
System
FIGURE 7.12 GIM structured approach. (Source: From Doumeingts et al., Methodologies for Designing CIM Systems:
A Survey, Figure 20, Elsevier Science, 1995. With permission.)
PERA
The Purdue Enterprise Reference Architecture (PERA) and its accompanying Purdue Methodology were
developed at Purdue University, (Nest Lafayette, IN). This university has also taken a leading role in the
definition of reference models for computer integrated manufacturing.
The Purdue Methodology is based on an extensive Instructional Manual that guides the preparation
of Master Plans. According to the methodology, an overall Master Plan is necessary before attempting to
implement any CIM program. A Master Plan includes a CIM Program Proposal (i.e., a set of integrated
projects whose completion will ensure the success of the desired integration of the enterprise). The Purdue
Enterprise Reference Architecture provides the framework for the development and use of the Instructional
Manual, the resulting Master Plan, and the ultimately implemented CIM Program Proposal. PERA is the
glue that holds together all aspects of the CIM program (Williams, 1994).
Figure 7.13 shows that the Purdue Enterprise Reference Architecture is characterized by the layered
structure of its life-cycle diagram. Starting with the Enterprise Business Entity, it leads to a description
of management’s mission, vision, and values for the entity under consideration. From these, the opera-
tional policies are derived for all elements of concern. Two—and only two—kinds of requirements are
derived from the policies, namely, those defining information-type tasks and physical manufacturing
tasks. Tasks become collected into modules or functions, which at their turn are connected into networks
of information or of material and energy flow. Then, technology choices are made; the role of the human
in the information architecture and in the manufacturing architecture is defined. The result of the technology
choices are three implementation architectures: namely, the information systems architecture, the human
and organizational architecture, and the manufacturing equipment architecture. After functional and
detailed design, the life cycle goes through the construction and operation phases, after which it is
disposed of (Williams, 1994).
A specific feature of PERA is the emphasis it puts on the role of humans. By defining the functions that
will be performed by humans, the information and manufacturing architectures are converted into the
information systems architecture, the human and organizational architecture, and the manufacturing
equipment architecture. Figure 7.14 illustrates that this definition involves three “lines.” The automatability
line shows the absolute extent of pure technological capability to automate tasks and functions, and is
limited by the fact that many tasks and functions require human innovation and cannot be automated with
presently available technology. The humanizability line shows the maximum extent to which humans can
be used to implement tasks and functions, and is limited by human abilities in speed of response, physical
strength, etc. The extent of automation line shows the actual degree of automation carried out or planned
in the system. This third line defines the boundary between the human and organizational architecture and
the information systems architecture on the one hand, and the boundary between the human and organi-
zational architecture and the manufacturing equipment architecture on the other side. Its location is
influenced by economic, political, social, and technological factors (Rathwell and Williams, 1996).
PERA explicitly takes into account the role of the human in a manufacturing system. In addition, it
does not distinguish between model content views such as function, information, and resource, but rather
between purpose views (information and manufacturing architecture) and implementation views
(human and organizational architecture, information systems architecture, and manufacturing equipment
architecture). The Purdue Methodology offers an enterprise engineering methodology that covers all
phases of a system life cycle. Of the three enterprise reference architectures described in this chapter,
PERA is clearly accompanied by the most extensive methodology. As for modeling languages, PERA offers
only the task module construct. Information system tasks, manufacturing tasks, and human-based tasks
Extent of automation
Extent of automation
of the manufacturing
Human component
Human component
of the information
Humanizability
Humanizability
Automatability
Automatability
architecture
architecture
FIGURE 7.14 Humanizability, automatability, and extent of automation to define the three implementation architectures.
(Source: From Rathwell, G.A. and Williams, T.J., Use of the Purdue Enterprise reference architecture and methodology
in industry (the Fluor Daniel example), in: Proceedings of the IFIP TC5 Working Conference on Models and Method-
ologies for Enterprise Integration, Bernus, P. and Nemes, L., Eds., pp. 12–44, 1996. With permission from Kluwer
Academic/Plenum Publishers.)
are modeled by means of this construct. By combining the various task modules into networks, a type
of data-flow or material and energy flow diagram results. There are no modeling methodologies known.
In addition, no modeling tool is known in literature. Although the only construct is quite simple,
models might become large and, without a tool, their accuracy over time is jeopardized. Earlier, the
developers of PERA had defined a reference model for computer-integrated manufacturing (Williams,
1989). This reference model is restricted to the automatable functions of an enterprise, and all functions that
might require human innovation are considered as external entities. PERA does not provide components of
an integrating infrastructure or any other generic enterprise modules.
GRAI/GIM and PERA provide some solutions to some of CIMOSA’s shortcomings. In particular, they
provide well-defined engineering methodologies: GRAI/GIM offers its structured approach, and PERA is
accompanied by the Purdue Methodology. For enterprise integration, however, it is necessary to keep an
up-to-date picture of an enterprise as time goes by. None of the three reference architectures can guarantee that.
Rules
................. Baan System
Configuration
.................
.................
Engineer-to-Order
Food Company X
Automotive
7.8 Summary
Reference architectures for enterprise integration aim to provide the necessary frameworks with which
companies might adapt their operation due to changing internal and external conditions. This chapter has
investigate the practical value of these enterprise reference architectures for the integration of production
systems. Three reference architectures have been studied; namely, CIMOSA, GRAI/GIM, and PERA. Their
overlaps and differences can be identified by means of the GERAM framework. Reference architectures should
at least be accompanied by engineering methodologies, modeling languages, and modeling methodologies. In
addition, modeling tools, reference models, and some generic enterprise modules are desirable components.
CIMOSA strives for the facilitation of continuous enterprise evolution. It intends to offer support for
business integration by means of its modeling framework, and for application integration by means of
its integrating infrastructure. Starting with the reference architecture and possible reference models, a
designer can make a CIMOSA model of its enterprise operations. The CIMOSA integrating infrastructure
processes the released model, so that operations are executed according to the model.
CIMOSA was applied during a reorganization project at Traub AG, a German tool manufacturer. The
modeling framework assisted Traub in the definition of a functional control architecture. Due to immature
specification of the integrating infrastructure, only the specification of the communication services was
helpful to Traub.
The CIMOSA reference architecture does not cover a full life cycle and lacks an accompanying engineering
methodology. However, CIMOSA does provide an eligible, although quite complex, framework for the
specification and analysis of functional architectures for production control systems. It prescribes how
to make a specification of the system, but it does not prescribe how to design the system. In addition, a
designer must personally make the translation from models to a real system. Therefore, CIMOSA should
be merely seen as a framework for the generation of documentation.
GRAI/GIM and PERA provide some solutions to some of CIMOSA’s shortcomings. In particular, they
provide well-defined engineering methodologies. For enterprise integration, however, it is necessary to
keep an up-to-date picture of an enterprise as time goes by, which is something the three reference
architectures cannot guarantee.
References
Aguiar, Marcos Wilson C., Ian Coutts, and Richard H. Weston. (1994). Model enactment as a basis for
rapid prototyping of manufacturing systems. In Proceedings of the European Workshop on Integrated
Manufacturing Systems Engineering (IMSE ’94), Grenoble, Dec. 1994, 86–96.
Aguiar, Marcos Wilson C., and Richard H. Weston. (1995). A model-driven approach to enterprise
integration. International Journal of Computer Integrated Manufacturing, Vol. 8, No. 3, pp. 210–224.
AMICE Consortium. (1993a). CIMOSA: Open System Architecture for CIM. Springer, Berlin.
AMICE Consortium. (1993b). CIMOSA: Open System Architecture for CIM—Technical Base Line, Version 2.0.
Bernus, Peter, Laszlo Nemes, and Theodore J. Williams. (1996). Architectures for Enterprise Integration.
Chapman & Hall, London.
Black, J.T. (1983). Cellular manufacturing systems reduce setup time, make small lot production economical.
Industrial Engineering, 15(11), 36–48.
Boudens, Gijs. (1997). Eenduidig opstellen van het business control model in de dynamic enterprise modeler.
M.Sc. thesis, Eindhoven University of Technology, Eindhoven (in Dutch).
Chen, D.G. Doumeingts, and L. Pun. (1990). An integrated inventory model based upon GRAI tools.
Engineering Costs and Production Economics, 19, 313–318.
Didic, M.M., F. Couffin, E. Holler, S. Lampérière, F. Neuscheler, J. Rogier, and M. de Vries. (1995). Open
engineering and operational environment for CIMOSA. Computers in Industry; Special Issue on
Validation of CIMOSA, 27(2), 167–178.
Doumeingts, Guy, Bruno Vallespir, Didier Darricau, and Michel Roboam. (1987). Design methodology
for advanced manufacturing systems. Computers in Industry, 9(4), 271–296.
Doumeingts, Guy, David Chen, and François Marcotte. (1992). Concepts, models and methods for the
design of production management systems. Computers in Industry, 19(1), 89–111.
Doumeingts, G., B. Vallespir, and D. Chen. (1995). Methodologies for designing CIM systems: a survey.
Computers in Industry; Special Issue on CIM in the Extended Enterprise, 25, 263–280.
Faure, J.M., A. Bassand, F. Couffin, and S. Lampérière. (1995). Business process engineering with partial
models. Computers in Industry; Special Issue on Validation of CIMOSA, 27(2), 111–122.
Gransier, Theo, and Werner Schönewolf. (1995). Editorial: Validation of CIMOSA. Computers in Industry;
Special Issue on Validation of CIMOSA, 27(2), 95–100.
Huizinga, Jan. (1995). Een gereedschapkist vol software-en businesskennis. Software Magazine, 12(10),
68–71 (in Dutch).
IFAC/IFIP Task Force on Architectures for Enterprise Integration. (1997). GERAM: Generalised Enterprise
Reference Architecture and Methodology. Available to download from http://www.cit.gu.edu.au/
~bernus/taskforce/geram/V1_2.html.
Janusz, Barbara. (1996). Modeling and reorganizing of process chains using CIMOSA. In Proceedings of
the IFIP TC5/WG5.3/WG5.7 International Conference on the Design of Information Infrastructure
Systems for Manufacturing (DIISM ’96), Jan Goossenaerts, Fumihiko Kimura, and Hans Wortmann,
Eds., Chapman & Hall, London, 140–151.
Keller, Gerhard, and Sören Detering. (1996). Process-oriented modeling and analysis of business processes
using the R/3 Reference Model. In Proceedings of the IFIP TC5 Working Conference on Models and
Methodologies for Enterprise Integration, Peter Bernus and Laszlo Nemes, Eds., Chapman & Hall,
London, 69–87.
Lapalus, Etienne, Shu Guei Fang, Christian Rang, and Ronald J. van Gerwen. (1995). Manufacturing
integration. Computers in Industry; Special Issue on Validation of CIMOSA, 27(2), 155–165.